The Routing Over Large Clouds Mailing List Archive by date

[Curtis Villamizar <curtis@ans.net>]

In message <199510251755.KAA19690@hubbub.cisco.com>, Yakov Rekhter writes:
> Curtis,
> 
> > I also don't see that we need a new term for describe a prefix which
> > covers a subset of the NBMA or covers the NBMA completely.  (Or
> > whatever APR was supposed to mean).
> 
> As you may remember, this document used the term LIS. However, at the
> IP-ATM WG meeting, when I presented this document it was quite clear
> consensus that a new term is needed, as the semantics (and constraints)
> of APR are different from an IP subnet (or LIS).
> 
> Yakov.


Yakov,

I remeber the following points made during that discussion:

  subnet - a small atomic unit that generally has a single prefix

    note: all hosts within a subnet prefix are considered directly
    reachable on the subnet.  There may be more than one prefix on a
    subnet, but a subnet prefix always exactly overlaps the subnet.
    (ignoring ugly proxy arp based hacks).

  network - one or more subnets connected by routing

    note:  network number is obsoleted by CIDR.  may have multiple
    prefixes as in the valid usage <org name's> network.

  internetwork - more than one network connected by routing

    generally assumed to have multiple prefixes

RFC1577 uses LIS rather than subnet so as not to confuse the OSI guys.

Currently network implies subnets connected by routing *only*.  The
notion of cut-through allows direct connections across subnets without
IP forwarding.

In an NBMA subnet (subnet often implied as in "for an NBMA OSPF
does..."), there is no broadcast, and all hosts are interconnected.
In a large NBMA network, there may be multiple NBMA subnets and from a
layer 2 standpoint all hosts are interconnected.

We need a term which defines the region for which cut-through or
direct connection is possible.  The term NBMA network fits well but
the term is defined as something that doesn't support broadcast so a
another term is wanted (maybe not needed, but wanted).

This is different from the set of prefixes covered by the NBMA.  I
don't remember a call for a new term to describe a prefix that falls
within an NBMA network.

Curtis